KR20200129397A - Method for controlling performance in virtualized environment and information processing device for the same - Google Patents

Abstract

Disclosed is an information processing device for performing performance control in virtualization environment. The information processing device comprises: an input/output thread that transmits signals between a virtual machine and at least one physical resource; and a performance management part that compares the target performance of the virtual machine and the actual performance of the virtual machine, and manages physical resources allocated to the input/output thread based on the comparison result, wherein the performance management part obtains information on the actual performance of the virtual machine, determines a performance adjustment value by using a difference between the actual performance of the virtual machine and the target performance of the virtual machine, and manages resources allocated to the input/output thread based on the performance adjustment value. According to the present invention, performance degradation due to resource contention between virtual machines can be prevented by controlling and managing input/output performance between a virtual machine and a physical device.

Description

Performance control method in a virtualized environment, and information processing device for the same {METHOD FOR CONTROLLING PERFORMANCE IN VIRTUALIZED ENVIRONMENT AND INFORMATION PROCESSING DEVICE FOR THE SAME}

The present invention relates to a performance control method in a virtualized environment and an information processing apparatus therefor.

A virtual machine (which can also be referred to as a VM, a virtual machine, a virtual machine, etc.) implements a physical computing environment (eg, a central processing unit or a memory device) in software to operate similarly to a general computing device. It is implemented so that it can be done. That is, the virtual machine is driven by using the resources of the physical machine. In the virtual machine, an operating system (OS, operating system, hereinafter referred to as guest operating system) for efficiently managing resources or software of the virtual machine may be installed. The guest operating system performs various operations, such as providing an interface to a predetermined resource, processing input data, or generating control commands for required operations, similar to an operating system directly installed on a physical machine (hereinafter referred to as a host operating system). can do.

A plurality of virtual machines may be installed and built in one physical machine. A plurality of virtual machines installed in one physical machine are driven while sharing resources of the physical machine. Virtualization is a technology that builds and provides an environment in which a number of virtual machines can share the resources of a physical machine while preventing a decrease in the resource utilization rate of a physical machine. The virtualization technology enables users to receive various computing platforms and services based on these platforms.

In recent years, cloud computing is also implemented by using such a virtualization technology. In particular, cloud computing may be implemented by virtualizing and consolidating physical resources using server consolidation technology. In the server integration technology, virtual machines installed on physical server devices with a relatively low utilization rate are centrally installed on a small number of different physical server devices, so that several virtual machines operate on one physical machine. Accordingly, it is possible to increase the utilization rate of a specific physical server device, and also reduce the maintenance cost of the entire cloud data center by cutting off power applied to other physical server devices that are not used. On the other hand, when a user uses a virtual machine constructed in this way, a certain level may be requested for the virtual machine according to the purpose of use. Cloud computing service providers are attempting to present and guarantee the level of user demand through a service level agreement (SLA). However, performance is still not properly guaranteed, and it remains at a limited level.

Republic of Korea Patent Registration No. 1613513 (2016.04.12.) Japanese Patent Publication No. 2018-036724 (2018.03.08.)

The present invention is to solve the provision of a performance control method in a virtualization environment and an information processing device for the same, which can prevent performance degradation due to resource contention between virtual machines by controlling and managing input/output performance between virtual machines and physical devices. It is a task to do.

In order to solve the above problems, a performance control method in a virtual environment and an information processing apparatus therefor are provided.

The information processing apparatus compares a virtual machine, an input/output thread for transmitting signals between the virtual machine and at least one physical resource, and a target performance of the virtual machine and the actual performance of the virtual machine, and the comparison result It may include a performance management unit for managing resources allocated to the input/output threads.

The performance management unit obtains information on the actual performance of the virtual machine, determines a performance adjustment value by using a difference between the actual performance of the virtual machine and the target performance of the virtual machine, and determines the performance adjustment value. Based on this, resources allocated to the input/output thread can be managed.

The performance management unit may determine a target usage of the input/output thread for the processor in response to the performance adjustment value, and manage resources of the processor allocated to the input/output thread based on the target usage for the processor.

The performance management unit may generate a control signal for at least one of the input/output thread and the processor based on the target usage.

The performance management unit calculates the performance adjustment value based on a ratio of the target performance to the difference between the actual performance and the target performance, and adds or subtracts the performance adjustment value to the current usage for the processor. The target usage can be determined.

The performance management unit may further use an adjustment constant to calculate the performance adjustment value.

The at least one physical resource may include at least one of a network device and a memory device.

The performance management unit may receive a target performance for the virtual machine from a user or another processor.

A performance control method in a virtual environment includes the steps of acquiring a target performance for a virtual machine and an actual performance for the virtual machine, and managing resources allocated to an input/output thread based on the target performance and the actual performance, the input/output thread May include transmitting a signal between the virtual machine and at least one physical resource.

The managing of the resources allocated to the input/output threads based on the target performance and the actual performance includes: obtaining information on the actual performance of the virtual machine, the actual performance of the virtual machine, and the target performance of the virtual machine. It may include determining a performance adjustment value using the difference between the values and managing resources allocated to the input/output thread based on the performance adjustment value.

The managing of the resources allocated to the input/output thread based on the performance adjustment value may include determining a target usage amount for the processor of the input/output thread in response to the performance adjustment value and the target usage for the processor. And managing resources of the processor allocated to the input/output thread.

Managing the resources allocated to the input/output thread based on the performance adjustment value may further include generating a control signal for at least one of the input/output thread and the processor based on the target usage.

Determining a performance adjustment value using the difference between the actual performance of the virtual machine and the target performance of the virtual machine, based on the ratio of the target performance to the difference between the actual performance and the target performance, the Comprising the step of calculating a performance adjustment value, and determining the target usage for the processor of the input/output thread in response to the performance adjustment value comprises: adding or subtracting the performance adjustment value from the current usage for the processor It may include the step of determining the target usage.

The calculating of the performance adjustment value may further include calculating the performance adjustment value by further using an adjustment constant.

The at least one physical resource may include at least one of a network device and a memory device.

The information processing device includes a plurality of virtual machines, a plurality of input/output threads corresponding to each of the plurality of virtual machines, and transmitting signals between the plurality of virtual machines and at least one physical resource, and a request of the plurality of virtual machines. A performance management unit that allocates resources to each of the plurality of input/output threads according to performance may be included.

According to the performance control method and the information processing device therefor in the above-described virtualization environment, it is possible to prevent performance degradation due to resource contention between virtual machines by controlling and managing input/output performance between virtual machines and physical devices, such as network devices or memory devices. You can get the effect that you can.

In addition, it is possible to efficiently use resources of a processor such as a central processing unit, and thus, it is possible to obtain an effect of additionally securing available resources of the processor.

In addition, it is possible to obtain an effect that a service provider of cloud computing can properly guarantee performance that meets the user's demand level.

In addition, it is possible to achieve the efficiency of operation and stabilization of the performance of the virtual machine without the installation or replacement of additional hardware equipment, thereby improving service quality and reducing costs.

A detailed description of each drawing is provided in order to more fully understand the drawings cited in the detailed description of the present invention.
1 is a diagram illustrating an embodiment of an information processing apparatus.
2 is a block diagram illustrating an embodiment of a performance management unit shown in FIG. 1.
FIG. 3 is a block diagram illustrating an embodiment of a performance monitoring unit and a performance control unit shown in FIG. 2.
FIG. 4 is a first diagram illustrating an operation before and after application of the performance management unit shown in FIG. 1.
FIG. 5 is a second diagram illustrating operations before and after application of the performance management unit shown in FIG. 1.
FIG. 6 is a flowchart illustrating a performance control method performed by the information processing device shown in FIG. 1 in a virtualization environment.

Hereinafter, the same reference numerals generally refer to the same components unless there is a special situation. In addition, a term to which'part' is added may be implemented as a software and/or hardware component, and according to an embodiment, a'part' may be implemented as a software and/or hardware component, or a'part' It is also possible for the part' to be implemented with a plurality of software and/or hardware components.

When a part is said to be connected to another part, it may mean a physical connection depending on a part and another part, or it may mean electrically connected. In addition, when a part includes another part, this does not exclude another part other than the other part unless otherwise stated, and it means that another part may be included further according to the designer's choice. do.

Terms such as'first' or'second' are used to distinguish one part from another, and unless otherwise specified, they do not mean sequential expressions. In addition, expressions in the singular may include plural expressions unless there is a clear exception in context.

Hereinafter, an embodiment of an information processing device capable of performing performance control in a virtualization environment will be described with reference to FIGS. 1 to 6.

1 is a diagram illustrating an embodiment of an information processing apparatus. Referring to FIG. 1, the information processing apparatus 1 may include hardware 50 and software 40 driven based on the hardware 50.

The hardware 50 refers to a physical device that is mounted on the information processing device 1 to perform an operation or supports the operation of the information processing device 1, and is required for software 40, for example, the host operating system 41, etc. Resources are provided so that the software 40 performs a predetermined operation. The hardware 50 may include a processor 51, a memory device 53, and a network device 55.

The processor 51 may generally or partially manage and control the information processing device 1 and perform arithmetic processing required for the operation of the information processing device 1 or software (programs, etc.) installed therein. The processor 51 may allocate and provide resources necessary for driving the host operating system 41 and/or at least one virtual machine 90 (90-1, 90-2) installed in the host operating system 41. . In addition, the processor 51 may allocate and provide necessary resources to at least one input/output thread 91 (91-1, 91-2) operating in the kernel region 42 of the host operating system 41.

The processor 51 is a central processing unit (CPU), a microcontroller unit (MCU), a microprocessor (Micom), an application processor (AP), an electronic control unit (ECU), Electronic Controlling Unit) and/or other electronic devices capable of processing various operations and generating control signals. These devices can be implemented using one or more semiconductor chips and related components.

The memory device 53 is provided to record and store data. The memory device 53 receives and writes data from the processor 51, the operating system 41, and/or various programs or virtual machines 90: 90-1, 90-2 operating within the operating system 41 It is also possible to provide the recorded data to these devices (41, 90, 51, etc.).

When the memory device 53 and the virtual machine (90: 90-1, 90-2) mutually transmit and receive requests, commands (eg, I/O processing requests, etc.), or data, the virtual machines 90: 90-1, 90-2) Input/output threads 91 (91-1, 91-2) corresponding to each may be used. That is, data of each virtual machine 90: 90-1, 90-2 may be transmitted to the memory device 53 through the input/output threads 91: 91-1, 91-2. Therefore, if sufficient resources (eg, central processing unit resources) are not allocated to the I/O threads 91: 91-1, 91-2, sufficient resources are provided for the virtual machines 90: 90-1, 90-2. Even if is allocated, data transmission/reception between the virtual machines 90 (90-1, 90-2) and the memory device 53 is liable to be delayed.

The memory device 53 may include at least one of a main memory device and an auxiliary memory device. Main memory may be implemented using ROM and/or RAM, and ROM is PROM, EPROM, EEPROM, and/or MASK-ROM. And the like, and the RAM includes DRAM and/or SRAM. Auxiliary memory devices include flash memory devices, SD (Secure Digital) cards, solid state drives (SSDs, Solid State Drives), hard disk drives (HDDs, Hard Disc Drives), magnetic drums, compact disks (CD), DVDs. Alternatively, it may be implemented using at least one storage medium capable of permanently or semi-permanently storing data, such as an optical media such as a laser disk, a magnetic tape, a magneto optical disk, and/or a floppy disk.

The network device 55 is provided to communicate with other information processing devices, external memory devices, and/or various devices that can be directly or indirectly connected to the other information processing device 1. The network device 55 may be implemented by assembling predetermined parts such as a communication chip, a communication antenna, a communication terminal, and/or other communication parts (eg, an amplifier, etc.).

According to an embodiment, the network device 55 uses the input/output threads 91: 91-1, 91-2 to correspond to the respective input/output threads 91: 91-1, 91-2. : 90-1, 90-2) and requests or commands (eg, I/O processing request, etc.) or data can be transmitted and received. Therefore, as described above, if appropriate resources are not allocated to each of the input/output threads 91: 91-1, 91-2, the network device 55 and the virtual machine 90: 90-1, 90-2 Task processing is also prone to delays.

The network device 55 can connect to a wired communication network, a wireless communication network, or a combination thereof, for example. Here, the wired communication network refers to a network built using a cable, and the cable may include, for example, a pair cable, a coaxial cable, an optical fiber cable or an Ethernet cable. The wireless communication network refers to a network established using at least one of a short-range communication network and a long-distance communication network. Short-distance communication networks include, for example, Wi-Fi, Wi-Fi Direct, Bluetooth, Bluetooth Low Energy, zigbee, and NFC (NFC, Near Field Communication) and/or CAN communication. The telecommunication network may be implemented based on a mobile communication standard such as a wired communication network, for example, 3GPP, 3GPP2, or WiMAX series.

In addition, the information processing device 1 may further include, for example, an input interface device such as a keyboard device or a mouse device, or an output interface device such as a monitor device or a speaker device.

The information processing device 1 described above may be implemented as at least one device capable of installing the virtual machines 90-1 and 90-2. For example, the information processing device 1 is a desktop computer, a laptop computer, a hardware device for a server, a network-attached storage device (NAS), a smart phone, a tablet PC, a head mounted display (HMD) device, a smart watch, Digital televisions, set-top boxes, navigation devices, personal digital assistants (PDAs), household appliances, mechanical devices, robots, vehicle control devices, portable game consoles, electronic boards, electronic billboards, sound reproduction devices and/or other virtual machines It may include at least one of various information processing devices that can be installed.

The software 40 may include an operating system 41 for managing, processing and/or controlling most processes performed by the information processing device 1. The operating system 41 may include, for example, Microsoft Windows, Mac OS, Linux, Unix, or various operating systems derived from Unix. At least one virtual machine (90: 90-1, 90-2) may be installed in the operating system 41, and the operating system 41 includes at least one virtual machine (90: 90-1, 90-2). ), it may be expressed as a host operating system.

The host operating system 41 may include a user area 43 and a kernel area 42. In the user area 43, various application programs (which can be expressed as applications or apps) that can be driven by the information processing device 1 may be executed. For example, each of the virtual machines 90: 90-1 and 90-2 may be driven in the user area 43 to provide various services to users.

The virtual machine (90: 90-1, 90-2) virtually implements a physical device (for example, the information processing device 1 itself or the physical parts 51, 53, 55, etc.) of the information processing device 1 In this case, a predetermined operating system (hereinafter referred to as a guest operating system) may be installed in the virtual machines 90: 90-1, 90-2 according to the user's or designer's selection. The operating system may be the same as or different from the host operating system 41. For example, the host operating system 41 may be a Linux operating system, and the guest operating system may be a Microsoft Windows operating system. One or a plurality of virtual machines 90: 90-1 and 90-2 may be installed in one host operating system 41. A plurality of virtual machines 90: 90-1 and 90-2 may be installed. In this case, each of the virtual machines 90: 90-1, 90-2 may operate like devices or operating systems that are physically separated from each other, in this case, a plurality of virtual machines 90: 90-1, 90-2 ) The same operating system may be installed on at least two of each other, or different operating systems may be installed.

According to an embodiment, the virtual machines 90: 90-1 and 90-2 may include a kernel-based virtual machine (KVM). The kernel-based virtual machine is a Linux-based virtualization platform, and can provide a virtualization environment in the form of a kernel module using Linux as the host operating system 41.

The kernel refers to software that manages and controls the system of the information processing device 1 as a whole. For example, the kernel is designed to perform various operations such as process management, resource management, or security. Such a kernel may include, for example, a single kernel, a micro kernel, a hybrid kernel, an exokernel, or a nanokernel. At least one kernel thread may be driven in the kernel region 42 implemented by the kernel. The kernel thread may include at least one input/output thread 91: 91-1, 91-2, and I/O thread.

I/O threads (91: 91-1, 91-2) transfer requests or data generated by virtual machines (90: 90-1, 90-2) to physical devices (51, 53, 55, etc.) or Requests or data from the devices 51, 53, 55, etc. may be transmitted to the virtual machines 90: 90-1, 90-2. The input/output threads 91: 91-1, 91-2 may be provided for each virtual machine 90: 90-1, 90-2 corresponding to each virtual machine 90: 90-1, 90-2. . For example, a first input/output thread 91-1 is provided to correspond to the first virtual machine 90-1, and a second input/output thread 91-2 corresponds to the second virtual machine 90-2. It can be prepared. Depending on the embodiment, one input/output thread 91 may be provided in a plurality of virtual machines 90, or a plurality of input/output threads 91 may be provided in one virtual machine 90. The input/output threads 91: 91-1 and 91-2 may operate in the above-described kernel-based virtual machine environment, but are not limited thereto.

According to an embodiment, the information processing apparatus 1 may include a performance management unit 100. Further, if necessary, the information processing apparatus 1 may further include a resource management unit (see 60 in Fig. 2).

The performance management unit 100 may be designed to manage the performance of one or more virtual machines 90 (90-1, 90-2) and/or perform various control operations necessary therefor. If a plurality of virtual machines (90: 90-1, 90-2) are installed in the host operating system (41), the performance management unit 100 is a plurality of virtual machines (90: 90-1, 90-2), respectively It can also be managed and controlled. As illustrated in FIG. 1, the performance management unit 100 may operate in the kernel region 42, and may be, for example, a kernel module operating in a virtualization environment. The kernel module may mean a code (program) that is loaded or unloaded in the kernel region 42 when necessary, and executed and terminated in the kernel region 42. Depending on the embodiment, the performance management unit 100 may be designed to operate in an area other than the kernel area 42. For example, the performance management unit 100 may be designed to perform a predetermined operation within the user area 41 or may be pre-embedded in a predetermined processor 51 to perform a predetermined operation.

According to an embodiment, the performance management unit 100 may manage the degree or weight (which can be expressed in terms of usage, occupancy, or occupancy) of the input/output threads 91: 91-1, 91-2. For example, the performance management unit 100 distributes and allocates resources of the processor 51 (for example, a central processing unit) provided to each of the input/output threads 91: 91-1 and 91-2. It is possible to manage, adjust, or control the usage of each of the at least one input/output thread 91: 91-1, 91-2 for resources. Accordingly, scheduling processing for managing the usage of resources of the input/output threads 91: 91-1, 91-2 becomes possible. The usage of the resources of the processor 51 may be expressed as a bandwidth of the central processing unit. Here, the bandwidth of the central processing unit is a value representing the usage of resources provided by the central processing unit and can be expressed as a percentage, and if one physical core is fully used, the bandwidth of the central processing unit is expressed as 100%. Can be.

The performance management unit 100 may be provided to operate in a kernel-based virtual machine environment. In this case, if the performance management unit 100 does not operate, the information processing device 1 operates in the same manner as a typical kernel-based virtual machine environment. Conversely, if the performance management unit 100 operates, the kernel-based In a virtual machine environment, it is possible to control the performance of the input/output threads 91: 91-1, 91-2, and accordingly, management of input/output performance for a physical device, for example, the memory device 53 or the network device 55, and Control becomes possible.

The performance management unit 100 may be implemented in the form of an application. In this case, the performance management unit 100 may be an application directly installed in the kernel area 42 by a designer or the like, or may be an application acquired through an electronic software distribution network accessible through a wired or wireless communication network.

FIG. 2 is a block diagram illustrating an embodiment of the performance management unit shown in FIG. 1, and FIG. 3 is a block diagram illustrating an embodiment of the performance monitoring unit and the performance control unit shown in FIG. 2.

According to the embodiment illustrated in FIG. 2, the performance management unit 100 may include a target performance receiver 110, a performance monitoring unit 130, and a performance control unit 150.

The target performance receiving unit 110 provides the performance required for each virtual machine (90-1 to 90-N, where N is a natural number greater than or equal to 1), and specific resources of each virtual machine (90-1 to 90-N). It is possible to receive and set the performance required for and/or the performance (hereinafter, target performance) required by each of the virtual machines 90-1 to 90-N. Here, the performance required for a specific resource of the virtual machines 90-1 to 90-N may include, for example, input/output performance for the memory device 53 and/or the network device 55 It may also include input/output capabilities for. In this case, the target performance for input/output of the memory device 53 can be input using a bandwidth (MB/s), and the target performance for input/output of the network device 55 is the number of packets processed per second or the bandwidth (Mbps). ) Can be used.

The target performance receiver 110 may receive a target performance from an input interface (not shown) or a processor 51. For example, the input interface may output an electrical signal according to a user's manipulation, and the target performance receiver 110 may receive the target performance by receiving the electrical signal output from the input interface or data corresponding thereto. The target performance receiver 110 may transmit the received target performance to the performance control unit 150.

The performance monitoring unit 130 collects information on the input/output threads 91-1 to 91-N corresponding to the virtual machines 90-1 to 90-N or the virtual machines 90-1 to 90-N, respectively. You may. For example, the performance monitoring unit 130 provides information on the amount of resources actually used by the virtual machines 90-1 to 90-N or the I/O threads 91-1 to 91-N (actual resource usage). It is possible to collect and acquire to monitor the operation of these (90-1 to 90-N, 91-1 to 91-N).

In addition, the performance monitoring unit 130 determines the actual performance (that is, the current performance) of the virtual machines 90-1 to 90-N or the input/output threads 91-1 to 91-N, and the determined actual performance It is also possible to determine whether or not performance control is necessary based on. To this end, the performance monitoring unit 130 may acquire performance information for the virtual machines 90-1 to 90-N or the input/output threads 91-1 to 91-N based on the collected information.

According to an embodiment, as shown in FIG. 3, the performance monitoring unit 130, in order to determine the actual performance and determine whether control is necessary, the control target selection unit 131, the performance information acquisition unit 133, and the target achievement It may include a determination unit 135.

The control target selection unit 131 includes at least one input/output thread to be controlled from among a plurality of input/output threads 91-1 to 91-N corresponding to each of the plurality of virtual machines 90-1 to 90-N, eg For example, the first input/output thread 91-1 may be selected. According to the embodiment, when a plurality of input/output threads 91-1 to 91-N exist, the control target selection unit 131 selects all of the plurality of input/output threads 91-1 to 91-N as control targets, respectively. Alternatively, only some of the input/output threads 91-1 to 91-N among the plurality of input/output threads 91-1 to 91-N may be selected as a control target. In addition, the control target selection unit 131 may sequentially or arbitrarily select all or part of the plurality of input/output threads 91-1 to 91 -N as a control target.

The performance information acquisition unit 133 acquires information necessary for the performance calculation of the virtual machines 90-1 to 90-N or the input/output threads 91-1 to 91-N, and based on this, the virtual machines 90- 1 to 90-N) or the actual performance of the input/output threads 91-1 to 91-N. In this case, the performance information acquisition unit 133 is a virtual machine corresponding to the first input/output thread 91-1 or the first input/output thread 91-1 selected by the control target selection unit 131 described above, that is, the first 1 It is possible to collect necessary information about the virtual machine 90-1 or to determine their actual performance. Here, the information required for the performance calculation may include, for example, the number of packets per second or bandwidth during the input/output operation of the network device 55 or the bandwidth during the input/output operation of the memory device 53. The performance information acquisition unit 133 may calculate real-time resource usage and actual performance by using predetermined statistical information about the resource management unit 60 and/or the device 1 separately provided to acquire the performance information.

The goal achievement determination unit 135 may receive a target performance from the target performance receiver 110 and compare the actual performance obtained by the performance information acquisition unit 133 with the target performance to determine whether the goal is achieved. Specifically, for example, the target achievement determination unit 135 calculates a difference between the actual performance and the target performance, and the difference between the actual performance and the target performance is equal to or within a predetermined range (for example, an error range). ), it can be determined that the goal has been achieved. Here, the predefined value may include 0. In other words, when the difference between the target performance and the actual performance is 0 (or the target performance and the actual performance are the same) or a value within a predefined range based on 0 (target performance-actual performance) Performance=0 or 0+a, where a is an arbitrary real number), virtual machines 90-1 to 90-N, or input/output threads 91-1 to 91-N may determine that the target performance has been reached. Here, a value within a predefined range may include, for example, a value approximating 0. Conversely, if the actual performance and the target performance are not 0 or have a value outside the predefined range (i.e., target performance-actual performance ≠ 0 or 0+a), the goal achievement determination unit 135 indicates that the goal has not been achieved. I can judge.

The result of determination of the performance monitoring unit 130 may be transmitted to the performance control unit 150, and information obtained by the performance monitoring unit 130 may be transmitted to the performance control unit 150 as necessary. For example, if it is determined that the target has not been achieved, the target achievement determination unit 135 may transmit the determination result that the target has not been achieved to the performance control unit 150, and in addition to the determination result, target performance, actual performance, and actual resource usage Information about the information may also be transmitted to the performance control unit 150. If necessary, the difference between the target performance calculated by the target achievement determination unit 135 and the actual performance may also be transmitted to the performance control unit 150. Conversely, if it is determined that the goal has been achieved, the goal achievement determination unit 135 may not transmit any determination result to the performance control unit 150, or, depending on the embodiment, transmit information on the goal achievement to the performance control unit 150. May be.

Collection of actual resource usage or acquisition of actual performance by the performance monitoring unit 130 may be performed periodically or aperiodically. For example, the performance monitoring unit 130 includes at least one input/output thread 91-1 to 91-N, for example, a selected first input/output thread 91- every predefined period (eg, 1 second). 1) resource usage can also be collected. Here, the resource usage of the first input/output thread 91-1 is the usage of the first input/output thread 91-1 for the processor 51 for performing communication with the memory device 53 and the usage of the network device 55 It may also include at least one of resource usage of the first input/output thread 91-1 for the processor 51 for performing communication.

The performance control unit 150 is based on the actual performance and goal achievement determination result derived by the performance monitoring unit 130, each of the virtual machines (90-1 to 90-N) or input/output threads (91-1 to 91-N) A performance adjustment value for is determined, and the actual usage of the input/output threads 91-1 to 91 -N may be controlled to be equal to or close to the target usage based on the performance adjustment value.

According to an embodiment, the performance control unit 150 may include a target usage determination unit 151 and a control signal generation unit 153.

The target usage determination unit 151 may perform resource scheduling adjustment by determining a target usage amount in response to reception of a signal from the performance monitoring unit 130. In other words, the target usage determination unit 151 determines that the actual performance of the first input/output thread 91-1 selected as the control target or the actual performance of the first virtual machine 90-1 corresponding thereto is required by the user or the like. When the target performance is not reached, the target of the first thread 91-1 for a predetermined resource so that the first input/output thread 91-1 or the first virtual machine 90-1 can achieve the target performance. You can decide how much to use. In this case, the target usage determination unit 151 may determine the target usage amount of the first thread 91-1 based on the actual performance, the target performance, and the actual usage.

Specifically, for example, the target usage determination unit 151 may determine the target usage (ie, resource usage) based on Equations 1 and 2 below.

은 성능 조절 값을 의미할 수 있다.Here, Ci denotes the actual usage of the resource (for example, the central processing unit) at a specific point in time (i), and Ci+1 may denote the target usage of the resource at the next point in time (i+1). . In addition, k is a value for adjusting the degree of change in the target usage amount (hereinafter, an adjustment constant) and has a real value of any one of a value between 0 and 1. If the adjustment constant (k) is relatively close to 0, the target usage is adjusted more precisely, and if the adjustment constant (k) is relatively close to 1, the target usage is adjusted to a larger width. The adjustment constant (k) may be given as 0.1, but this is exemplary, and the adjustment constant (k) can be variously defined according to a designer or user's selection. Pg means the quantified target performance, and Pr means the quantified actual performance. The target performance Pg may be obtained by the target performance receiver 110 and the actual performance Pr may be obtained by the performance information acquisition unit 133. On the right side of Equation 1

May mean a performance adjustment value.

Equation 1 described above may be rewritten as Equation 2 below.

Equation 2 is for converting a performance adjustment value for achieving a target performance into a CPU bandwidth parameter. Here, Qi means the available time of the resource (eg, the central processing unit) at a specific point in time (i), Qi+1 means the available time of the resource at the next point (i+1), and F is It means the period (us). For example, when the available time (Qi) of the resource is 10,000 μs and the period (F) is 100,000 μs, the bandwidth of the resource allocated to the selected predetermined input/output thread 91-1 is 10% (=10,000/100,000 * 100).

As described in Equations 1 and 2, the target usage determination unit 151 divides the difference between the target performance (Pg) and the actual performance (Pr) at a specific time point (i) by the target performance to adjust the performance. By acquiring the value and adding the performance adjustment value to the actual usage (Ci) at a specific point in time (i), it is possible to determine the target usage (Ci+1) for the selected target (i.e., the first I/O thread 91-1). have. In this case, the target usage determination unit 151 obtains a performance adjustment value by adding an adjustment constant k to a value obtained by dividing the difference between the target performance and the actual performance by the target performance, and calculates the obtained performance adjustment value as the actual usage ( It is also possible to calculate the target usage amount (Ci+1) at the next time point (i+1) after weighting in addition to Ci).

When determining the target usage (Ci+1) as described above, if the collected actual performance (Pr) is lower than the target performance (Pg) (i.e., Pr <Pg), the performance adjustment value is positive (+). It has a value, and the target usage amount (Ci+1) for the selected first input/output thread 91-1 is determined to be higher than the actual usage amount (Ci). In other words, it is determined to increase the resource usage of the first input/output thread 91-1. Conversely, if the collected actual performance (Pr) is higher than the target performance (Pg) (ie, Pr> Pg), the above-described performance adjustment value has a negative (-) value, and accordingly, the first input/output thread 91 The target usage (Ci+1) for -1) is determined to be lower than the actual usage (Ci). In other words, it is determined to reduce the resource usage of the first input/output thread 91-1.

According to an embodiment, the target usage determination unit 151 calculates an absolute value of the difference between the target performance Pg and the actual performance Pr at a specific point in time (i) and divides the absolute value by the target performance to obtain a performance adjustment value. You can also get it. In this case, the target usage determination unit 151 first compares the size between the actual performance (Pr) and the target performance (Pg), and if the actual performance (Pr) is lower than the target performance (Pg) (ie, Pr < Pg), the target usage (Ci+1) is calculated by adding the performance adjustment value to the actual usage (Ci). Conversely, if the actual performance (Pr) is higher than the target performance (Pg) (i.e., Pr> Pg), the performance adjustment The target usage amount (Ci+1) can also be calculated by subtracting the value from the actual usage amount (Ci).

The determination result of the target usage determination unit 151 may be transmitted to the control signal generation unit 153. The control signal generation unit 153 generates a control signal for an increase or decrease in usage in response to the determination result of the target usage determination unit 151, and the input/output thread selected based on this, that is, the first input/output thread 91-1 ) Resource usage can be controlled. That is, the resource allocated to the first input/output thread 91-1 varies according to the operation of the control signal generator 153. The control signal is transmitted, for example, to the processor 51 or the virtual machines 90-1 to 90-N, and thus the resource usage of the input/output thread 91-1 increases or decreases. According to an embodiment, the control signal generation unit 153 may control the resource usage of the first input/output thread 91-1 by using the resource management unit 60. In this case, the control signal is transmitted to the resource management unit 60, and the resource management unit 60 may control resources allocated to the first input/output thread 91-1 in response thereto.

The resource management unit 60 provides information necessary for determining resource usage or performance information to the performance monitoring unit 130 and/or the virtual machines 90-1 to 90-N according to the control of the performance control unit 150 Alternatively, resources allocated to the input/output threads 91-1 to 91-N corresponding thereto may be controlled. The resource management unit 60 may be provided to limit or isolate the use of resources of processes or threads, and may be implemented as a kernel module operating in the kernel region 42. According to an embodiment, the resource management unit 60 may be implemented using a control group (cGroup) provided by a Linux operating system or the like, but is not limited thereto.

The operations of the performance management unit 100 described above include a plurality of I/O threads 91-1 to 90-N corresponding to each of the plurality of virtual machines 90-1 to 90-N or the plurality of virtual machines 90-1 to 90-N. 91-N), and sequentially or randomly for all of the plurality of virtual machines 90-1 to 90-N or all of the plurality of I/O threads 91-1 to 91-N Alternatively, it may be performed sequentially or arbitrarily on a part of the plurality of virtual machines 90-1 to 90-N or a part of the plurality of input/output threads 91-1 to 91-N. Also, the operation of the performance management unit 100 described above may be repeatedly performed periodically or aperiodically. For example, the performance of the first virtual machine 90-1 and/or the first I/O thread 91-1 may be monitored periodically or aperiodically, and the resource usage of the first virtual machine 90-1 may be It may change.

As described above, when the actual performance does not reach the target performance while the virtual machine 90: 90-1 to 90-N performs various tasks, the virtual machine 91: 90-1 to 90-N Resources (eg, the bandwidth of the central processing unit) provided to each input/output thread (91: 91-1 to 91-N) can be adjusted, and accordingly, the performance required by the user or the like can be achieved more quickly and This can be done easily. In this way, the process of adjusting resource allocation through feedback control can be performed in real time without prior profiling, so that overhead for pre-profiling can be minimized.

Hereinafter, the performance of the information processing device 1 when the above-described performance management unit 100 does not operate and the performance of the information processing device 1 when the performance management unit 100 operates are compared with each other to compare the performance management unit ( The effect of 100) will be described in detail.

FIG. 4 is a first diagram illustrating an operation before and after application of the performance management unit shown in FIG. 1. FIG. 4 is to allow the first virtual machine (VM1) and the second virtual machine (VM2) to perform a network operation (consecutive transmission operation of 1024B TCP packets) under an environment in which the performance management unit 100 does not operate and an operation environment. It shows the measurement result of throughput according to the performance. In FIG. 4, the y-axis represents throughput (unit: pps), and the x-axis represents a case in which the performance management unit 100 does not operate or operates sequentially from the left.

Referring to FIG. 4, when the performance management unit 100 does not operate, both the first virtual machine VM1 and the second virtual machine VM2 operate at a performance of approximately 2000 pps in the same manner. On the contrary, when the performance management unit 100 is operated and the target performance of the first virtual machine VM1 is set to 10,000 pps and the target performance of the second virtual machine VM2 is set to 25,000 pps, the first virtual machine ( The VM1) and the second virtual machine VM2 operate with differentiated performance according to the set target performance. Specifically, the first virtual machine (VM1) operates with a lower performance (10,000pps) in response to the target performance lower than the actual performance, and the second virtual machine (VM2) corresponds to the target performance higher than the actual performance. It operates with higher performance (25,000pps).

FIG. 5 is a second diagram illustrating operations before and after application of the performance management unit shown in FIG. 1. 5 shows a first virtual machine (VM1) and a second virtual machine (VM2) in an environment in which the performance management unit 100 does not operate and performs randomly writing and reading operations of a 16kB file in a storage device. (Disk operation) shows the result of measuring the bandwidth. The y-axis of FIG. 4 represents a bandwidth (unit: MBps), and the x-axis represents a case in which the performance management unit 100 does not operate and operates sequentially from the left.

Referring to FIG. 5, when the performance management unit 100 does not operate, both the first virtual machine VM1 and the second virtual machine VM2 operated with the same performance (approximately 220 MBps), but the performance management unit If (100) is operated and the target performance of the first virtual machine (VM1) is set to 100 MBps and the target performance of the second virtual machine (VM2) is set to 300 MBps, the first virtual machine (VM1) and the second virtual machine ( It can be seen that each VM2) operates differently according to the set target performance. Specifically, the first virtual machine (VM1) operates with a performance corresponding to a target performance relatively low compared to the actual performance (100 MBps), and the second virtual machine (VM2) operates at a target performance that is relatively high compared to the actual performance. It can be seen that it operates with performance (300MBps).

As shown in FIGS. 4 and 5, the virtual machines VM1 and VM2 operate differently from each other according to the operation of the performance management unit 100 and the differential setting of the target performance, and accordingly, the virtual machine VM1 , VM2) resources can be properly distributed.

Hereinafter, an embodiment of a performance control method in a virtualization environment will be described with reference to FIG. 6.

FIG. 6 is a flowchart illustrating a performance control method performed by the information processing device shown in FIG. 1 in a virtualization environment.

As shown in FIG. 6, first, an input/output thread as a control target may be selected (401). In this case, the selected input/output thread may be any one input/output thread among a plurality of input/output threads, for example, an i-th input/output thread (i is a natural number of 1 or more).

The actual performance of the input/output thread or the virtual machine corresponding to the input/output thread as a control target, and the resource usage of the input/output thread or the virtual machine may be obtained (403). In addition, target performance may also be obtained before, after or simultaneously with the acquisition of actual performance and usage. The target performance may be input by a user, may be predefined by a designer, or may be determined separately by an information processing device.

The target performance and the actual performance may be sequentially compared (405). If the target performance and the actual performance are the same (YES in 405), the adjustment value calculation or target usage-based control processes 407 to 415 to be described later may not be performed. Depending on the embodiment, even when only the difference between the target performance and the actual performance is a constant value (eg, an error range), it may be determined that the performance is substantially the same. In this case, the adjustment value calculation or target usage-based control processes 407 to 415 described later may not be performed.

If the target performance and the actual performance are different (NO in 405), a performance adjustment value may be calculated accordingly (407). The performance adjustment value may be a value obtained by dividing the difference between the target performance and the actual performance by the target performance, as described in Equations 1 and 2, or obtained by multiplying the obtained value by an adjustment constant. It can also be a value. In this case, the performance adjustment value may have a positive value or a negative value according to the target performance and the actual performance. If the target performance is greater than the actual performance, the performance adjustment value may have a positive value. Conversely, if the actual performance is greater than the target performance, the performance adjustment value may have a negative value. The performance adjustment value given in this way is added to the actual usage, and the target usage can be obtained according to the result.

According to an embodiment, the performance adjustment value may be an absolute value obtained by dividing the absolute value of the difference between the target performance and the actual performance by the target performance or dividing the difference between the target performance and the actual performance by the target performance. In this case, the target performance and the actual performance are first compared (409).

If the target performance is greater than the actual performance (YES in 409), the target usage may be determined by adding the performance adjustment value to the actual usage (411). Conversely, if the target performance is less than the actual performance (NO of 409), the target usage may be determined by subtracting the performance adjustment value from the actual usage (413).

In either process, if the target performance is greater than the actual performance, the target usage is determined to be larger than the actual usage, and conversely, when the target performance is smaller than the actual performance, the target usage is determined to be smaller than the actual usage.

When the target usage is determined, the usage of the input/output thread i for the resource (eg, the central processing unit) may be changed based on the target usage (415). For example, the resources allocated to the selected input/output thread i may be increased or decreased. In this case, resource allocation for input/output threads may be controlled using a resource management interface separately installed in the kernel or the like.

Depending on the embodiment, whether to control input/output threads other than the sequentially selected input/output threads (i) or whether to additionally control the selected input/output threads (i) may be determined (417). If it is determined whether to perform additional control (YES in 417), a predetermined input/output thread (i+k, k ≥ 0) is selected, and the above-described processes 401 to 417 may be repeated. Here, k may be 1 or may be a natural number other than 1. In other words, all of the plurality of input/output threads (i+k, k ≥ 0) may be sequentially selected according to a predefined or random order, or among a plurality of input/output threads (i+k, k ≥ 0) Only some input/output threads (eg, input/output threads (i+2, i=1, 3, 5)) may be selected. Depending on the embodiment, it is also possible to select the same input/output thread (i, k=0) again.

The performance control method according to the above-described embodiment may be implemented in the form of a program that can be driven by a computer device. Here, the program may include a program command, a data file, a data structure, or the like alone or in combination. The program may be designed and produced using machine code or high-level language code. The program may be specially designed to implement the performance control method in the above-described virtual environment, or may be implemented using various functions or definitions that are known and available to those skilled in the computer software field. In addition, the computer device may be implemented including a processor or a memory that enables the function of a program to be realized, and may further include a communication device if necessary.

A program for implementing the above-described performance control method may be recorded in a computer-readable recording medium. The computer-readable recording medium includes, for example, a magnetic disk storage medium such as a hard disk or a floppy disk, a magnetic tape, an optical recording medium such as a compact disk or a DVD, a magnetic-optical recording medium such as a floppy disk, and a ROM. , A semiconductor storage device such as RAM or flash memory, etc., may include various types of hardware devices capable of storing a specific program executed according to a call from a computer.

The performance control method in a virtualization environment and an example of an information processing device for the same have been described above, but the performance control method in a virtualization environment and an information processing device therefor are not limited to the above-described embodiments. Various devices or methods that can be implemented by modifying and transforming a person skilled in the art based on the above-described embodiments may also be examples of the performance control method and apparatus therefor in the above-described virtualization environment. For example, the described techniques are performed in a different order from the described method, and/or components such as the described system, structure, device, circuit, etc. are combined, combined, or described in a form different from the described method, Performance control method in the above-described virtual environment even if other components are additionally added to components such as structures, devices, and circuits, or components such as described systems, structures, devices, and circuits are replaced or replaced by other components And it may be an embodiment of a device for this.

1: information processing unit 40: software
41: host operating system 50: hardware
51: processor 53: memory device
55: network device 60: resource management unit
90: virtual machine 91: I/O thread
100: performance management unit 110: target performance receiver
130: performance monitoring unit 150: performance control unit

Claims (14)

An input/output thread for transferring a signal between the virtual machine and at least one physical resource; And
A performance management unit that compares the target performance of the virtual machine and the actual performance of the virtual machine, and manages physical resources allocated to the input/output threads based on the comparison result,
The performance management unit obtains information on the actual performance of the virtual machine, determines a performance adjustment value using the difference between the actual performance of the virtual machine and the target performance of the virtual machine, and determines a performance adjustment value based on the performance adjustment value. To manage the resources allocated to the input/output thread,
Information processing device.
The method of claim 1,
The performance management unit determines a target usage amount for the processor of the input/output thread in response to the performance adjustment value and manages resources of the processor allocated to the input/output thread based on the target usage for the processor,
Information processing device.
The method of claim 2,
The performance management unit generates a control signal for at least one of the input/output thread and the processor based on the target usage,
Information processing device.
The method of claim 2,
The performance management unit calculates the performance adjustment value based on the ratio of the target performance to the difference between the actual performance and the target performance, adds or subtracts the performance adjustment value to the current usage for the processor, To determine the target usage,
Information processing device.
The method of claim 4,
The performance management unit further uses an adjustment constant to calculate the performance adjustment value,
Information processing device.
The method of claim 1,
The at least one physical resource includes at least one of a network device and a memory device.
The method of claim 1,
The performance management unit receives a target performance for the virtual machine from a user or another processor,
Information processing device.
In a virtualization environment, in the performance control method performed by the information processing device,
Obtaining a target performance for a virtual machine and an actual performance for the virtual machine; And
And managing resources allocated to an input/output thread based on the target performance and the actual performance, wherein the input/output thread transfers a signal between the virtual machine and at least one physical resource.
The method of claim 8,
Managing resources allocated to input/output threads based on the target performance and the actual performance,
Obtaining information on the actual performance of the virtual machine;
Determining a performance adjustment value using a difference between the actual performance of the virtual machine and the target performance of the virtual machine; And
And managing resources allocated to the input/output thread based on the performance adjustment value,
How to control performance.
The method of claim 9,
Managing resources allocated to the input/output thread based on the performance adjustment value,
Determining a target usage amount of the input/output thread for the processor in response to the performance adjustment value; And
Including the step of managing resources of the processor allocated to the input/output thread based on the target usage for the processor,
How to control performance.
The method of claim 10,
Managing resources allocated to the input/output thread based on the performance adjustment value,
Generating a control signal for at least one of the input/output thread and the processor based on the target usage, further comprising,
How to control performance.
The method of claim 10,
Determining a performance adjustment value using the difference between the actual performance of the virtual machine and the target performance of the virtual machine, based on the ratio of the target performance to the difference between the actual performance and the target performance, the Computing a performance adjustment value,
Determining the target usage for the processor of the input/output thread in response to the performance adjustment value includes determining the target usage by adding or subtracting the performance adjustment value from the current usage for the processor,
How to control performance.
The method of claim 12,
The step of calculating the performance adjustment value further comprises calculating the performance adjustment value by further using an adjustment constant,
How to control performance.
The method of claim 8,
The at least one physical resource includes at least one of a network device and a memory device,
How to control performance.

Images